Collaborative Research:CISE-MSI:RCBP-RF:CNS:Orchestration of Network Slicing for 5G-Enabled IoT Devices Using Reinforcement Learning

无线通信是从一台设备向另一台设备传输信息的最重要的媒介之一。目前的大多数无线电话都支持4G或5G网络。5G旨在为多个用户提供更快的数据速度、更高的可用性和统一的用户体验。5G先进能力将影响医疗、教育、娱乐、物联网(IoT)、自动驾驶汽车和智能城市等多个行业。这项研究旨在创建一种能够有效管理连接到5G网络的物联网设备的系统。管理具有不同需求的大量物联网设备是一项复杂的任务，这使得手动管理具有挑战性。一些设备要求观看视频或玩虚拟现实游戏等活动的快速数据传输，而另一些设备则要求自动驾驶汽车或监控设备等任务的快速响应时间。这些问题的解决方案是网络切片，它涉及将网络划分为更小的部分，以处理不同类型的设备和服务。然而，网络切片所固有的挑战是高效地管理网络资源、协调和优化网络的不同部分。该项目通过设计一个可以自动管理5G物联网设备资源的系统来应对这些挑战。此方法的潜在好处是简化了网络、降低了成本、节省了能源、平衡了工作负载、优化了移动性，并使网络更易于管理。这项研究为研究5G物联网网络中的机器学习和网络自动化奠定了坚实的基础，从而推动了该领域的发展。此外，通过聘用和指导STEM中代表性不足的学生，该项目将旨在弥合美国各地院校之间的差距。该项目将培训来自少数群体服务大学和边缘化社区的下一代学者，并帮助5G和强化学习(RL)领域的劳动力发展。项目负责人还将接触K-12，以促进教育，并与包括女性在内的不同范围的学生接触。该项目的目标是设计一个框架，用于自动化支持5G的物联网设备的端到端资源管理，该框架在大规模网络中利用RL技术和大规模多输入多输出(MIMO)。5G网络中各种用例、设备和应用的多样化需求使得手动操作成本高、难度大、效率低。该项目将考虑敏捷性，以确保网络能够快速适应不断变化的需求。它旨在降低网络复杂性和成本，节约网络能量，优化负载平衡和移动性，并简化资源管理。研究的范围是a)使用海量MIMO为物联网设备设计5G网络切片，b)开发RL模型来解决大规模5G网络中的物联网设备协调问题，c)将RL解决方案集成到大规模MIMO网络切片的5G支持物联网网络中。5G网络切片方法将根据每个切片的特定需求将资源分配到每个切片，并通过采用大规模MIMO技术和RL模型最大限度地减少运营成本(OPEX)和资本支出(CAPEX)来提供网络即服务。这一方法将为支持RL的大规模MIMO 5G网络带来更高的可用性、指定的延迟、更快的速度、更好的安全性和更高的吞吐量。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Wireless communication is one of the most important mediums for transmitting information from one device to another. Most of the current wireless phones are supported by either 4G or 5G networks. 5G is meant to deliver higher data speeds, increased availability, and a uniform user experience to multiple users. 5G advanced capabilities will impact several industries including healthcare, education, entertainment, Internet of Things (IoT), autonomous vehicles, and smart cities. This research aims to create a system that can effectively manage IoT devices connected to the 5G network. Managing a multitude of IoT devices with diverse requirements is a complex task, making manual management challenging. Some devices require fast data transmission for activities like watching videos or playing virtual-reality games, while others need a quick response time for tasks like self-driving cars or monitoring devices. The solution to these problems is network slicing which involves dividing the network into smaller parts to handle different types of devices and services. However, the challenges inherent to network slicing are efficiently managing network resources, coordinating, and optimizing different parts of the network. This project addresses these challenges by designing a system that can automatically manage the resources of 5G-enabled IoT devices. The potential benefits of this approach are that it simplifies the network and reduces cost, saves energy, balances the workload, optimizes mobility, and makes the network easier to manage. This research advances the field by laying a solid groundwork for studying machine learning and network automation in devices that are part of the 5G-enabled IoT network. Furthermore, by employing and mentoring students from underrepresented backgrounds in STEM, this project will aim to bridge the gap in institutions across the US. This project will train the next generation of scholars from minority-serving universities and marginalized communities and help in workforce development in the fields of 5G and reinforcement learning (RL). The project leaders will also reach out to K-12 to promote education and engage with a diverse range of students, including women.The goal of this project is to devise a framework for automating end-to-end resource management of 5G-enabled IoT devices that utilizes RL techniques with massive multiple-input multiple-output (MIMO) in large-scale networks. The diverse needs of various use cases, devices, and applications in 5G networks make manual operation costly, difficult, and inefficient. This project will consider agility to ensure that the network can quickly adapt to evolving requirements. It aims to decrease network complexity and cost, conserve network energy, optimize load balancing and mobility, and simplify resource management. The scope of the research is a) designing 5G network slicing using Massive MIMO for IoT devices, b) developing an RL model to solve orchestration problems of IoT devices in large-scale 5G networks, and c) integrating the RL solution into a Massive MIMO network sliced 5G-enabled IoT network. The 5G network-slicing approach will enable resource allocation to each slice considering its specific needs and provide networks-as-a-service by minimizing operational expenses (OPEX) and capital expenditure (CAPEX) by adopting the Massive MIMO technique and RL models. This approach will result in higher availability, a specified latency, faster speed, better security, and higher throughput of RL-enabled Massive MIMO 5G networks.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.